De Robbio, Antonella Predire l’imprevedibile: perché l’IA ha sempre smentito i suoi profeti., 2025 [Preprint]
Preview |
Text
2025PredireIlBoLive.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (529kB) | Preview |
English abstract
This paper explores the epistemological and structural limitations of predictive artificial intelligence when dealing with the complexity and unpredictability of the real world. The article highlights how machine learning models and generative algorithms, relying heavily on historical training data, are inherently limited when forecasting rare events, systemic shifts, or chaotic behaviors ("black swans"). By analyzing the crucial distinction between statistical correlation and actual causation, the author argues that data-driven induction alone cannot replace theoretical scientific modeling and abductive reasoning. The paper concludes by emphasizing the urgent need to integrate AI tools with robust theoretical scientific frameworks, positioning human expertise and critical intuition as an indispensable "safety net" for the conscious and ethical application of predictive technologies.
Italian abstract
Il presente contributo indaga i limiti epistemologici e strutturali dell'intelligenza artificiale predittiva di fronte alla complessità e all'imprevedibilità del mondo reale. L'articolo evidenzia come i modelli di machine learning e gli algoritmi generativi, essendo addestrati su serie storiche e dati del passato, siano intrinsecamente limitati nel prevedere eventi rari, rotture sistemiche o comportamenti caotici (i "cigni neri"). Esplorando la differenza cruciale tra correlazione statistica e reale nesso di causalità, il testo sottolinea come la pura induzione basata sui dati non possa sostituire la formulazione teorica e il ragionamento abduttivo. L'autrice conclude richiamando la necessità di integrare l'AI con la solida conoscenza scientifica teorica, ponendo l'esperienza e l'intuito critico umano come indispensabile ancoraggio per un uso consapevole ed etico delle tecnologie predittive.
| Item type: | Preprint |
|---|---|
| Keywords: | Artificial intelligence; Predictive models; Complex systems; Information theory; Causality and correlation; Algorithmic limits; Epistemology of science; Black swans; Machine learning; Human oversight. |
| Subjects: | A. Theoretical and general aspects of libraries and information. > AB. Information theory and library theory. B. Information use and sociology of information > BC. Information in society. L. Information technology and library technology > LP. Intelligent agents. |
| Depositing user: | Antonella De Robbio |
| Date deposited: | 15 Jul 2026 13:30 |
| Last modified: | 15 Jul 2026 13:30 |
| URI: | http://hdl.handle.net/10760/47903 |
Downloads
Downloads per month over past year
Actions (login required)
![]() |
View Item |
